Documentation·Get Started
Get Started

Introduction

SenateLab turns a single coding CLI into a team of coordinated AI agents that share a workspace, a memory, and a goal. This page covers what the product does, why it exists, and how the pieces fit together.

What is SenateLab?

SenateLab is a desktop hub for teams of AI agents. It installs locally on macOS, Windows, or Linux, and it drives whichever Codex CLI session you already use. Instead of flipping between terminal tabs, every agent lives on a shared spatial map — you can see who is active, who is waiting, and what has already shipped. The canvas stays open in one window, so the whole team's state is always a glance away, not a tmux scavenger hunt.

Why SenateLab?

  • Teams beat solos. A single strong agent handles one task. Five focused agents with a clear handoff protocol handle five in parallel, because they narrate blockers to each other instead of to you.
  • Spatial beats tabs. Terminal tabs hide state. A visual canvas makes idle time, contention, and waiting-for-you moments obvious the moment they happen.
  • Isolation by folder. Each agent is scoped to an explicit set of folders. A QA agent cannot touch production config. A frontend agent cannot rewrite your Terraform.
  • Memory that compounds. Every agent keeps a persistent report — decisions, gotchas, what worked and what did not — so the third time you hit the same bug, someone on the team already has notes on it.

How it works

You connect a Codex CLI install once. From there, SenateLab spawns agents on top of that engine — each gets its own session, folder scope, and memory store. The canvas is the control surface: drag to move, click to open, type to ask. Watch-mode keeps a chat live while the agent works, and task badges tell you at a glance how much is in flight.

1

Connect your coding CLI

Install Codex CLI once and point SenateLab at the binary. Every agent uses that same engine — no per-agent auth dance.

2

Spawn agents with a role and a scope

Pick a role, grant folder access, write a short purpose. The agent lands on your canvas with a name and an identity.

3

Watch the team work

Tasks, messages, and status updates all live on one canvas. Jump in when an agent is waiting on you; stay out of the way when it isn't.

  • One Codex CLI install powers every agent. No per-agent authentication dance.
  • Agents run as isolated sessions; nothing leaks between them by accident.
  • The canvas updates in real time — status ring, sticky note, animated message line.
  • Pause anyone with Esc and hand off to another agent without losing context.

Next steps